HuggingGPT: Ordering the Subtasks
Advanced
The problem, from first principles
HuggingGPT (Shen et al., 2023) uses a language model to plan a set of subtasks, each handled by a specialist model. The subtasks form a dependency graph, and they must run in an order where each task's inputs are ready.
From theory to code
Implement topo_order(tasks), returning an execution order consistent with the dependencies, or None if the plan has a cycle.
Constraints
- Break ties by task id so the order is deterministic.
Hints
Hint 1
Repeatedly run a task whose dependencies are all finished. Track how many dependencies each task still waits on, and stop with None if some task never becomes ready.
Keywords
research-papersagentsplanningorchestration
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.